664 missing assertions for threaded execution

[byjtew]

Closes #278

[anyzelman]

Ready to merge

[GiovaGa]

@anyzelman this seems solved, I don't know if we want a clearer error message to be outputted instead of a possibly cryptic assertion error.
I will run unit tests on the internal CI just to be sure, but should be good to go

[anyzelman]

I rebased again, reviewed it, and running smoke + unit tests with LPF now (in addition to CI).

Changes during review: I made the check a function of the Coordinates class, as otherwise it is not easy to call for a check without a lot of code repetition. I also modified the internal::getCoordinates for const-coordinates to perform the check. I did not think the modification to config::OMP was needed and reverted that, instead relying on its already-established threads function call.

There is one annoyance that in the use of the coordinates class, a callee tends to call a third class for creating the buffer, and then passes that into Coordinates. This means that the number of threads is actually controlled by this third class. The current design in this MR is hence not totally robust if somehow the number of threads spawned during a parallel section changes between the time the third party class code executes and allocates the buffer for some T1 #threads , and the time the #threads is actually recorded when the coordinates instance is set while some other T2 #threads is active.